NSW Agriculture, Centre for Sheep Meat Development, P.O. Box 129, Cowra, NSW 2794, Australia.
Meat Sci. 2004 Jun;67(2):269-74. doi: 10.1016/j.meatsci.2003.10.015.
A wide selection of lamb types of mixed sex (ewes and wethers) were slaughtered at a commercial abattoir and during this process images of 360 carcasses were obtained online using the VIAScan® system developed by Meat and Livestock Australia. Soft tissue depth at the GR site (thickness of tissue over the 12th rib 110 mm from the midline) was measured by an abattoir employee using the AUS-MEAT sheep probe (PGR). Another measure of this thickness was taken in the chiller using a GR knife (NGR). Each carcass was subsequently broken down to a range of trimmed boneless retail cuts and the lean meat yield determined. The current industry model for predicting meat yield uses hot carcass weight (HCW) and tissue depth at the GR site. A low level of accuracy and precision was found when HCW and PGR were used to predict lean meat yield (R(2)=0.19, r.s.d.=2.80%), which could be improved markedly when PGR was replaced by NGR (R(2)=0.41, r.s.d.=2.39%). If the GR measures were replaced by 8 VIAScan® measures then greater prediction accuracy could be achieved (R(2)=0.52, r.s.d.=2.17%). A similar result was achieved when the model was based on principal components (PCs) computed from the 8 VIAScan® measures (R(2)=0.52, r.s.d.=2.17%). The use of PCs also improved the stability of the model compared to a regression model based on HCW and NGR. The transportability of the models was tested by randomly dividing the data set and comparing coefficients and the level of accuracy and precision. Those models based on PCs were superior to those based on regression. It is demonstrated that with the appropriate modeling the VIAScan® system offers a workable method for predicting lean meat yield automatically.
从一个商业屠宰场屠宰了大量不同性别的绵羊（母羊和公羊），在此过程中，使用澳大利亚肉类和牲畜协会（Meat and Livestock Australia）开发的 VIAScan®系统在线获取了 360 具尸体的图像。屠宰场员工使用 AUS-MEAT 绵羊探头（PGR）测量 GR 部位的软组织深度（距中线 12 肋 110 毫米处的组织厚度）。在冷藏室中使用 GR 刀（NGR）测量了该厚度的另一个度量值。随后，每具尸体被分解为一系列修剪后的无骨零售切块，并确定瘦肉产量。目前，用于预测肉产量的行业模型使用热胴体重（HCW）和 GR 部位的组织深度。当使用 HCW 和 PGR 预测瘦肉产量时，发现准确性和精密度水平较低（R(2)=0.19，r.s.d.=2.80%），当用 NGR 替代 PGR 时，准确性和精密度可以显著提高（R(2)=0.41，r.s.d.=2.39%）。如果用 8 个 VIAScan®测量值替代 GR 测量值，则可以实现更高的预测准确性（R(2)=0.52，r.s.d.=2.17%）。当模型基于从 8 个 VIAScan®测量值计算的主成分（PCs）时，也可以达到类似的结果（R(2)=0.52，r.s.d.=2.17%）。与基于 HCW 和 NGR 的回归模型相比，使用 PCs 还提高了模型的稳定性。通过随机划分数据集并比较系数以及准确性和精密度，测试了模型的可移植性。基于 PCs 的模型优于基于回归的模型。证明了在适当的建模下，VIAScan®系统为自动预测瘦肉产量提供了一种可行的方法。
